Cleanup of includes directory. (#246)

Now breaking includes into directories. Co-authored-by: jluitjens <jluitjens@nvidia.com>
NVIDIA · Aug 15, 2022 · 0c5c8df · 0c5c8df
1 parent 1281272
commit 0c5c8df
Show file tree

Hide file tree

Showing 121 changed files with 2,069 additions and 1,449 deletions.
diff --git a/bench/00_transform/einsum.cu b/bench/00_transform/einsum.cu
@@ -1,6 +1,5 @@
 #include <nvbench/nvbench.cuh>
 #include "matx.h"
-#include "matx_einsum.h"
 
 #if ENABLE_CUTENSOR
 

diff --git a/examples/convolution.cu b/examples/convolution.cu
@@ -30,9 +30,7 @@
 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /////////////////////////////////////////////////////////////////////////////////
 
-#include "matx_type_utils.h"
-#include "matx_tensor.h"
-#include "matx_conv.h"
+#include "matx.h"
 #include <cassert>
 #include <cstdio>
 

diff --git a/examples/fft_conv.cu b/examples/fft_conv.cu
@@ -31,7 +31,6 @@
 /////////////////////////////////////////////////////////////////////////////////
 
 #include "matx.h"
-#include "matx_filter.h"
 #include <cassert>
 #include <cstdio>
 #include <cuda/std/ccomplex>

diff --git a/examples/recursive_filter.cu b/examples/recursive_filter.cu
@@ -31,7 +31,6 @@
 /////////////////////////////////////////////////////////////////////////////////
 
 #include "matx.h"
-#include "matx_filter.h"
 #include <cassert>
 #include <cstdio>
 #include <cuda/std/ccomplex>

diff --git a/examples/spectrogram.cu b/examples/spectrogram.cu
@@ -30,14 +30,13 @@
 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /////////////////////////////////////////////////////////////////////////////////
 
-#include "matx.h"
-#include "matx_viz.h"
-#include "matx_make.h"
 #include <cassert>
 #include <cstdio>
 #include <math.h>
 #include <memory>
 
+#include "matx.h"
+
 using namespace matx;
 #define FFT_TYPE CUFFT_C2C
 
@@ -156,4 +155,4 @@ int main([[maybe_unused]] int argc, [[maybe_unused]] char **argv)
   cudaStreamDestroy(stream);
   CUDA_CHECK_LAST_ERROR();
   MATX_EXIT_HANDLER();
-}
+}
diff --git a/examples/spectrogram_graph.cu b/examples/spectrogram_graph.cu
@@ -31,7 +31,6 @@
 /////////////////////////////////////////////////////////////////////////////////
 
 #include "matx.h"
-#include "matx_viz.h"
 #include <cassert>
 #include <cstdio>
 #include <math.h>
@@ -165,4 +164,4 @@ int main([[maybe_unused]] int argc, [[maybe_unused]] char **argv)
   cudaStreamDestroy(stream);
   CUDA_CHECK_LAST_ERROR();
   MATX_EXIT_HANDLER();
-}
+}
diff --git a/include/matx.h b/include/matx.h
@@ -32,28 +32,19 @@
 
 #pragma once
 #include <cuda/std/ccomplex>
-#include "matx_defines.h"
-#include "matx_half_complex.h"
-#include "matx_half.h"
+#include "matx/core/defines.h"
+#include "matx/core/error.h"
+#include "matx/core/file_io.h"
+#include "matx/core/half_complex.h"
+#include "matx/core/half.h"
+#include "matx/core/pybind.h"
+#include "matx/core/tensor.h"
+#include "matx/core/utils.h"
 
-#include "matx_utils.h"
-#include "matx_error.h"
-#include "matx_tensor.h"
-#include "matx_random.h"
-#include "matx_tensor_generators.h"
+#include "matx/executors/executors.h"
+#include "matx/generators/generators.h"
 #include "matx/operators/operators.h"
 #include "matx/transforms/transforms.h"
-#include "matx_exec_kernel.h"
-#include "matx_fft.h"
-#include "matx_conv.h"
-#include "matx_corr.h"
-#include "matx_matmul.h"
-#include "matx_reduce.h"
-#include "matx_inverse.h"
-#include "matx_solver.h"
-#include "matx_cov.h"
-#include "matx_cub.h"
-
 
 using fcomplex = cuda::std::complex<float>;
 using dcomplex = cuda::std::complex<double>;

diff --git a/include/matx_allocator.h → include/matx/core/allocator.h b/include/matx_allocator.h → include/matx/core/allocator.h
@@ -41,7 +41,7 @@
 #include <cuda_runtime_api.h>
 #endif
 
-#include "matx_error.h"
+#include "matx/core/error.h"
 
 #pragma once
 

diff --git a/include/matx_cache.h → include/matx/core/cache.h b/include/matx_cache.h → include/matx/core/cache.h
@@ -33,11 +33,12 @@
 
 #pragma once
 
-#include "matx_error.h"
 #include <functional>
 #include <optional>
 #include <unordered_map>
 
+#include "matx/core/error.h"
+
 namespace matx {
 namespace detail {
 

diff --git a/include/matx_defines.h → include/matx/core/defines.h b/include/matx_defines.h → include/matx/core/defines.h
diff --git a/include/matx_error.h → include/matx/core/error.h b/include/matx_error.h → include/matx/core/error.h
@@ -37,7 +37,7 @@
 #include <exception>
 #include <sstream>
 
-#include "matx_stacktrace.h"
+#include "matx/core/stacktrace.h"
 
 namespace matx
 {

diff --git a/include/matx_file_io.h → include/matx/core/file_io.h b/include/matx_file_io.h → include/matx/core/file_io.h
@@ -30,8 +30,6 @@
 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 /////////////////////////////////////////////////////////////////////////////////
 
-#include "matx_pybind.h"
-#include "matx_tensor.h"
 #include <cstdio>
 #include <iterator>
 #include <shared_mutex>
@@ -41,7 +39,9 @@
 // #include <cudf/io/csv.hpp>
 // #include <cudf/types.hpp>
 
-#include "matx_error.h"
+#include "matx/core/pybind.h"
+#include "matx/core/tensor.h"
+#include "matx/core/error.h"
 
 #pragma once
 
@@ -231,4 +231,4 @@ void WriteMAT(const TensorType &t, const std::string fname,
 }
 
 }; // namespace io
-}; // namespace matx
+}; // namespace matx
diff --git a/include/matx_get_grid_dims.h → include/matx/core/get_grid_dims.h b/include/matx_get_grid_dims.h → include/matx/core/get_grid_dims.h
diff --git a/include/matx_half.h → include/matx/core/half.h b/include/matx_half.h → include/matx/core/half.h
@@ -32,12 +32,13 @@
 
 #pragma once
 
-#include "cuda_bf16.h"
-#include "cuda_fp16.h"
 #include <cuda/std/cmath>
-#include "matx_defines.h"
 #include <type_traits>
 
+#include "cuda_bf16.h"
+#include "cuda_fp16.h"
+#include "matx/core/defines.h"
+
 namespace matx {
 
 /**
@@ -1209,4 +1210,4 @@ __MATX_HOST__ __MATX_DEVICE__ __MATX_INLINE__ matxHalf<T> tanh(const matxHalf<T>
 using matxFp16 = matxHalf<__half>; ///< Alias for fp16
 using matxBf16 = matxHalf<__nv_bfloat16>; ///< Alias for bf16
 
-}; // namespace matx
+}; // namespace matx
diff --git a/include/matx_half_complex.h → include/matx/core/half_complex.h b/include/matx_half_complex.h → include/matx/core/half_complex.h
@@ -33,13 +33,14 @@
 #pragma once
 // Next line is a workaround for libcuda++ 11.6.0 on CTK 11.4.100+. Remove once libcuda++ is fixed
 #define _LIBCUDACXX_HAS_NO_INT128
-#include "cuComplex.h"
-#include "matx_half.h"
 #include <complex>
 #include <cuda/std/complex>
 #include <cuda/std/cmath>
 #include <type_traits>
 
+#include "cuComplex.h"
+#include "matx/core/half.h"
+
 namespace matx {
 
 /**
@@ -1043,4 +1044,4 @@ tanh(const matxHalfComplex<T> &x)
 using matxFp16Complex = matxHalfComplex<matxFp16>; ///< Alias for a MatX fp16 complex wrapper
 using matxBf16Complex = matxHalfComplex<matxBf16>; ///< Alias for a MatXbf16 complex wrapper
 
-}; // namespace matx
+}; // namespace matx
diff --git a/include/matx_make.h → include/matx/core/make_tensor.h b/include/matx_make.h → include/matx/core/make_tensor.h
@@ -33,9 +33,9 @@
 /////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include "matx_storage.h"
-#include "matx_tensor_desc.h"
-#include "matx_shape.h"
+#include "matx/core/storage.h"
+#include "matx/core/tensor_desc.h"
+#include "matx/core/shape.h"
 
 namespace matx {
 
@@ -353,4 +353,4 @@ auto make_static_tensor() {
   return tensor_t<T, desc.Rank(), decltype(s), decltype(desc)>{std::move(s), std::move(desc)};
 }
 
-} // namespace matx
+} // namespace matx
diff --git a/include/matx_pybind.h → include/matx/core/pybind.h b/include/matx_pybind.h → include/matx/core/pybind.h
@@ -31,7 +31,7 @@
 /////////////////////////////////////////////////////////////////////////////////
 #pragma once
 
-#include "matx.h"
+#include "matx/core/type_utils.h"
 
 
 #include <pybind11/embed.h>

diff --git a/include/matx_shape.h → include/matx/core/shape.h b/include/matx_shape.h → include/matx/core/shape.h
@@ -36,9 +36,9 @@
 #include <iomanip>
 #include <type_traits>
 
-#include "matx_allocator.h"
-#include "matx_error.h"
-#include "matx_type_utils.h"
+#include "matx/core/allocator.h"
+#include "matx/core/error.h"
+#include "matx/core/type_utils.h"
 
 namespace matx {
 

diff --git a/include/matx_stacktrace.h → include/matx/core/stacktrace.h b/include/matx_stacktrace.h → include/matx/core/stacktrace.h
diff --git a/include/matx_storage.h → include/matx/core/storage.h b/include/matx_storage.h → include/matx/core/storage.h
@@ -32,9 +32,9 @@
 
 #pragma once
 
-#include "matx_type_utils.h"
-#include "matx_allocator.h"
-#include "matx_error.h"
+#include "matx/core/type_utils.h"
+#include "matx/core/allocator.h"
+#include "matx/core/error.h"
 
 namespace matx
 {
@@ -538,4 +538,4 @@ namespace matx
  */
   template <typename T>
   using DefaultStorage = basic_storage<raw_pointer_buffer<T, owning, matx_allocator<T>>>;
-};
+};
diff --git a/include/matx_tensor.h → include/matx/core/tensor.h b/include/matx_tensor.h → include/matx/core/tensor.h
@@ -40,12 +40,12 @@
 #include <memory>
 #include <type_traits>
 
-#include "matx_allocator.h"
-#include "matx_error.h"
-#include "matx_storage.h"
-#include "matx_tensor_impl.h"
-#include "matx_utility_kernels.cuh"
-#include "matx_tensor_utils.h"
+#include "matx/core/allocator.h"
+#include "matx/core/error.h"
+#include "matx/core/storage.h"
+#include "matx/core/tensor_impl.h"
+#include "matx/core/tensor_utils.h"
+#include "matx/kernels/utility.cuh"
 
 static constexpr int MAX_TENSOR_DIM = 4;
 

diff --git a/include/matx_tensor_desc.h → include/matx/core/tensor_desc.h b/include/matx_tensor_desc.h → include/matx/core/tensor_desc.h
@@ -36,7 +36,7 @@
 
 #include <array>
 #include <type_traits>
-#include "matx_error.h"
+#include "matx/core/error.h"
 
 namespace matx {
 
@@ -445,4 +445,4 @@ template <int RANK>
 using DefaultDescriptor = tensor_desc_cr_ds_64_64_t<RANK>;
 
 
-}; // namespace matx
+}; // namespace matx
diff --git a/include/matx_tensor_impl.h → include/matx/core/tensor_impl.h b/include/matx_tensor_impl.h → include/matx/core/tensor_impl.h
@@ -34,15 +34,15 @@
 
 #include <type_traits>
 #include <cuda/std/functional>
-#include "matx_error.h"
-#include "matx_set.h"
-#include "matx_defines.h"
-#include "matx_tensor_desc.h"
-#include "matx_type_utils.h"
-#include "matx_tensor_utils.h"
-#include "matx_exec_kernel.h"
-#include "matx_executor.h"
-#include "matx_make.h"
+#include "matx/core/error.h"
+#include "matx/core/defines.h"
+#include "matx/core/tensor_desc.h"
+#include "matx/core/type_utils.h"
+#include "matx/core/tensor_utils.h"
+#include "matx/operators/set.h"
+//#include "matx_exec_kernel.h"
+//#include "matx_executor.h"
+#include "matx/core/make_tensor.h"
 
 namespace matx {
 

diff --git a/include/matx_tensor_utils.h → include/matx/core/tensor_utils.h b/include/matx_tensor_utils.h → include/matx/core/tensor_utils.h
@@ -34,7 +34,8 @@
 
 #include <cuda/std/tuple>
 #include <functional>
-#include "matx_make.h"
+
+#include "matx/core/make_tensor.h"
 
 namespace matx
 {
@@ -618,4 +619,4 @@ void Print(const Op &op, Args... dims) {
 //     }, tup);
 // }   
 
-}
+}
diff --git a/include/matx_type_utils.h → include/matx/core/type_utils.h b/include/matx_type_utils.h → include/matx/core/type_utils.h
@@ -32,9 +32,6 @@
 
 #pragma once
 
-#include "cuda_fp16.h"
-#include "matx_half.h"
-#include "matx_half_complex.h"
 #include <any>
 #include <memory>
 #include <complex>
@@ -43,6 +40,10 @@
 #include <cuda/std/tuple>
 #include <type_traits>
 
+#include "cuda_fp16.h"
+#include "matx/core/half.h"
+#include "matx/core/half_complex.h"
+
 /**
  * Defines type traits for host and device compilers. This file should be includable by
  * the host compiler, so no device code should be present

diff --git a/include/matx_utils.h → include/matx/core/utils.h b/include/matx_utils.h → include/matx/core/utils.h
@@ -31,8 +31,8 @@
 /////////////////////////////////////////////////////////////////////////////////
 
 #pragma once
-#include "matx_defines.h"
-#include "matx_error.h"
+#include "matx/core/defines.h"
+#include "matx/core/error.h"
 
 #define AMPERE_CC 8
 #define VOLTA_CC 7